Document Understanding on AI2D (test)
88.9AccuracyQwen3-VL 32B
Evaluation results
| Method | Links | |
|---|---|---|
| Qwen3-VL 32BModel Source Category=Open-weight VLMs, Model Parameters=32B, Evaluation Mode=Thinking Mode2026.01 | 88.9 | |
| Gemini-2.5 FlashModel Source Category=Closed-source VLMs, Evaluation Mode=Thinking Mode2026.01 | 88.7 | |
| GPT5 miniModel Source Category=Closed-source VLMs, Evaluation Mode=Thinking Mode2026.01 | 88.2 | |
| MMFineReason-8BModel Source Category=Ours, Model Parameters=8B, Evaluation Mode=Thinking Mode2026.01 | 87.9 | |
| Qwen3-VL 30B-A3BModel Source Category=Open-weight VLMs, Model Parameters=30B-A3B, Evaluation Mode=Thinking Mode2026.01 | 86.9 | |
| MMFineReason-4BModel Source Category=Ours, Model Parameters=4B, Evaluation Mode=Thinking Mode2026.01 | 86.5 | |
| HoneyBee 8BModel Source Category=Open-source VLMs, Model Parameters=8B, Evaluation Mode=Thinking Mode2026.01 | 86 | |
| OMR 7BModel Source Category=Open-source VLMs, Model Parameters=7B, Evaluation Mode=Thinking Mode2026.01 | 85 | |
| Qwen3-VL 8BModel Source Category=Open-weight VLMs, Model Parameters=8B, Evaluation Mode=Thinking Mode2026.01 | 84.9 | |
| MMR1 8BModel Source Category=Open-source VLMs, Model Parameters=8B, Evaluation Mode=Thinking Mode2026.01 | 83.4 | |
| MMFineReason-2BModel Source Category=Ours, Model Parameters=2B, Evaluation Mode=Thinking Mode2026.01 | 82.5 | |
| MRoPEPositional Encoding Method=MRoPE, DIPE Enhancement=+DIPE2026.03 | 67.39 | |
| MRoPE-IPositional Encoding Method=MRoPE-I, DIPE Enhancement=+DIPE2026.03 | 67.39 | |
| Vanilla RoPEPositional Encoding Method=Vanilla RoPE, DIPE Enhancement=+DIPE2026.03 | 66.48 | |
| MRoPE-IPositional Encoding Method=MRoPE-I, DIPE Enhancement=Base2026.03 | 66.26 | |
| MRoPEPositional Encoding Method=MRoPE, DIPE Enhancement=Base2026.03 | 65.32 | |
| Vanilla RoPEPositional Encoding Method=Vanilla RoPE, DIPE Enhancement=Base2026.03 | 65.06 |